Perpustakaan Digital ITB

This thesis quantifies the risk of aircraft hard landings using a physics-based model and subset simulation on operational flight data. Hard landings, although rare, contribute significantly to landing-related incidents in aviation. Traditional statistical methods often fail to capture the true risk of such rare events due to limited occurrence in datasets. To address this, a simplified physical model was developed to estimate vertical acceleration (????????????????) during landing using Quick Access Recorder (QAR) data from 4,204 Boeing 747 flights at KMSP airport. The dataset was preprocessed by isolating the landing phase (1000 ft AGL to 15 seconds after touchdown), classifying runways, and smoothing sensor noise. The model treats the aircraft as a rigid body and incorporates parameters such as pitch angle, angle of attack, elevator deflection, engine N1, true airspeed (reconstructed from groundspeed and wind), and barometric altitude. A hard landing is defined as exceeding 1.8g at touchdown. Model outputs were validated against QAR data and used in subset simulation to estimate the probability of a hard landing. Ten independent subset simulation runs (4000 total samples, conditional level 0.1) yielded a failure probability of ~3.04125 × 10?7 with high statistical stability. A sensitivity analysis using ±5% perturbations revealed ground speed as the most influential parameter, followed by pitch and angle of attack. These findings highlight that hard landing risk is dominated by parameters controlling vertical energy state, emphasizing the importance of precise speed control and pilot technique during flare. The method offers a robust approach for assessing rare-event risks in aviation safety.